Krugman picks up a point that I have also recently followed in a Robert Gordon paper, namely:
"there’s an important distinction between the prices of wheat, oil, rubber, etc. that may rise or fall by double-digit amounts over the course of a year, then quickly reverse that rise or fall, and the prices of many services and manufactured goods — and most wages — which are set for periods of months or years. The latter are slow to develop inflation, but also slow to give it up, which is why policy should focus on whether those prices have started to rise too fast (or too slowly).
The question, however, is whether changes in the flexible-price goods feed into persistent inflation in the core. Phelps thought not: he believed that wages were set mainly in reference to other wages, implying that swings in oil or wheat prices were largely irrelevant to the story. I’d agree: if we think of wages as the ultimate core price, I don’t see any mechanism in today’s America whereby rising commodity prices translate into higher wage contracts."
Some prices are highly flexible- especially those that are set in auction type markets. Others are sticky owing to contractual relations- most notably concerning labor. What this says is that policy makers should pay attention to core inflation in setting monetary policy, and ignore seasonal fluctuations in auction markets. Especially now that unions have lost clout, and COLA contracts are less common.
This is at odds with some Economists who claim that the key measure to watch is the headline inflation number. Of course, a sub-issue here is how the core or trimmed inflation figure is put together. When is core-inflation really core-inflation versus a mere statistical squib? If we are simply knocking off the top and bottom 15% of price movements in a given period is this really portraying "core" inflation accurately?